Social Explorative Attention Networks (SEAN)

This repository provides a reference implementation of sean as described in the paper in SIGKDD 2019:

Beyond Personalization: Social Content Recommendation for Creator Equality and Consumer Satisfaction.
Wenyi Xiao, Huan Zhao, Haojie Pan, Yangqiu Song, Vincent W. Zheng, Qiang Yang.

You are welcome to conduct your research based on our work, and please kindly cite our paper with the following bib file:

@inproceedings{Xiao2019Sean,
author = {Xiao, Wenyi and Zhao, Huan and Pan, Haojie and Song, Yangqiu and Zheng, Vincent W. and Yang, Qiang},
title = {Beyond Personalization: Social Content Recommendation for Creator Equality and Consumer Satisfaction},
booktitle = {SIGKDD},
year = {2019}}

The sean algorithm goes beyond personalized content recommendation by considering both content creators and consumers, which motivates us to develop a highly personalized attention based model and explore higher-order social friends.

Basic Usage

Dataset download

English Files down load: please put files in the dir dataset/steemit/en/

processed_user_activity.json

https://1drv.ms/u/s!AorJby8-9jo1gy6SzwIpwEs2DdaM

new_article.json

https://1drv.ms/u/s!AorJby8-9jo1gy2MJPREepy6Sgjq

Example

To run sean on Steemit-En, you can use the following command:
cd sean

python steemit_preprocessing

python payout.py --walk-length 10 --num-walks 3 --alpha 1

Input

The supported input format is an edgelist:

node1_id_int node2_id_int

The graph is assumed to be directed and unweighted by default.

Output

The probability of clicking an unseen document by the target user.

Miscellaneous

Please send any questions you might have about the code and/or the algorithm to wxiaoae@cse.ust.hk.